relation: https://eprints.untirta.ac.id/58484/
title: ANALISIS SENTIMEN TWITTER TERHADAP APLIKASI MYTELKOMSEL MENGGUNAKAN MODEL BIDIRECTIONAL LONG SHORT-TERM MEMORY (BILSTM)
creator: JULKIFLI, JULKIFLI
subject: QA75 Electronic computers. Computer science
subject: QA76 Computer software
subject: T Technology (General)
description: User satisfaction evaluation of the MyTelkomsel application is often hindered by the limited scope of conventional survey methods, which fail to capture real-time user responses. This study aims to analyze national-scale public sentiment using a Deep Learning method with Bidirectional Long Short-Term Memory (Bi-LSTM) architecture. Review data was collected via web scraping techniques using Tweet Harvest software and automatically labeled through a Transformer-based Ensemble Learning approach (IndoBERT, RoBERTa, and BERT-SMSA) to address the lack of labels in raw data. The developed classification model integrates FastText Word Embedding features to handle informal language characteristics and out-of vocabulary (OOV) words dominant in social media. Testing results using the Stratified 5-Fold Cross Validation method showed that the proposed model achieved an average accuracy of 96%, with an F1-Score of 93% on the negative class. This study concludes that the combination of Bi-LSTM architecture and automatic labeling techniques proves effective in accurately and consistently representing user sentiment as a strategic reference for service development.
date: 2026-02-18
type: Thesis
type: NonPeerReviewed
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/1/Julkifli_3337210026_Fulltext.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/2/Julkifli_3337210026_01.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/3/Julkifli_3337210026_02.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/5/Julkifli_3337210026_03.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/6/Julkifli_3337210026_04.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/7/Julkifli_3337210026_05.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/8/Julkifli_3337210026_Ref.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/9/Julkifli_3337210026_Lamp.pdf
format: text
language: id
identifier: https://eprints.untirta.ac.id/58484/10/Julkifli_3337210026_CP.pdf
identifier:   JULKIFLI, JULKIFLI  (2026) ANALISIS SENTIMEN TWITTER TERHADAP APLIKASI MYTELKOMSEL MENGGUNAKAN MODEL BIDIRECTIONAL LONG SHORT-TERM MEMORY (BILSTM).  S1 thesis, Fakultas Teknik Universitas Sultan Ageng Tirtayasa.